def from_data(cls, data, col_variable, row_variable=None):
if row_variable is None:
row_variable = data.domain.class_var
if row_variable is None:
raise ValueError("row_variable needs to be specified (data"
"has no class)")
row_variable = _get_variable(row_variable, data, "row_variable")
col_variable = _get_variable(col_variable, data, "col_variable")
try:
dist, unknowns = data._compute_contingency(
[col_variable], row_variable)[0]
self = super().__new__(cls, dist.shape)
self[...] = dist
self.unknowns = unknowns
except NotImplementedError:
self = np.zeros(
(len(row_variable.values), len(col_variable.values)))
self.unknowns = 0
rind = data.domain.index(row_variable)
cind = data.domain.index(col_variable)
for row in data:
rval, cval = row[rind], row[cind]
if math.isnan(rval):
continue
w = row.weight
if math.isnan(cval):
self.unknowns[cval] += w
else:
self[rval, cval] += w
self.row_variable = row_variable
self.col_variable = col_variable
return self
def from_data(cls, data, col_variable, row_variable=None):
if row_variable is None:
row_variable = data.domain.class_var
if row_variable is None:
raise ValueError("row_variable needs to be specified (data "
"has no class)")
row_variable = _get_variable(row_variable, data, "row_variable")
col_variable = _get_variable(col_variable, data, "col_variable")
try:
dist, unknowns = data._compute_contingency([col_variable],
row_variable)[0]
self = super().__new__(cls, dist.shape)
self[...] = dist
self.unknowns = unknowns
except NotImplementedError:
self = np.zeros(
(len(row_variable.values), len(col_variable.values)))
self.unknowns = 0
rind = data.domain.index(row_variable)
cind = data.domain.index(col_variable)
for row in data:
rval, cval = row[rind], row[cind]
if math.isnan(rval):
continue
w = row.weight
if math.isnan(cval):
self.unknowns[cval] += w
else:
self[rval, cval] += w
self.row_variable = row_variable
self.col_variable = col_variable
return self
def from_data(self, data, col_variable, row_variable=None):
if row_variable is None:
row_variable = data.domain.class_var
if row_variable is None:
raise ValueError("row_variable needs to be specified (data" "has no class)")
self.row_variable = _get_variable(row_variable, data, "row_variable")
self.col_variable = _get_variable(col_variable, data, "col_variable")
try:
(self.values, self.counts), self.unknowns = data._compute_contingency([col_variable], row_variable)[0]
except NotImplementedError:
raise NotImplementedError("Fallback method for computation of " "contingencies is not implemented yet")
def from_data(self, data, col_variable, row_variable=None):
if row_variable is None:
row_variable = data.domain.class_var
if row_variable is None:
raise ValueError("row_variable needs to be specified (data"
"has no class)")
self.row_variable = _get_variable(row_variable, data, "row_variable")
self.col_variable = _get_variable(col_variable, data, "col_variable")
try:
self[:], self.unknowns = data._compute_contingency([col_variable],
row_variable)[0]
except NotImplementedError:
raise NotImplementedError("Fallback method for computation of "
"contingencies is not implemented yet")
def from_data(cls, data, col_variable, row_variable=None):
if row_variable is None:
row_variable = data.domain.class_var
if row_variable is None:
raise ValueError(
"row_variable needs to be specified (data has no class)")
row_variable = _get_variable(row_variable, data, "row_variable")
col_variable = _get_variable(col_variable, data, "col_variable")
try:
dist, col_unknowns, row_unknowns, unknowns = \
data._compute_contingency([col_variable], row_variable)[0]
self = super().__new__(cls, dist.shape)
self[...] = dist
self.col_unknowns = col_unknowns
self.row_unknowns = row_unknowns
self.unknowns = unknowns
except NotImplementedError:
shape = len(row_variable.values), len(col_variable.values)
self = super().__new__(cls, shape)
self[...] = np.zeros(shape)
self.col_unknowns = np.zeros(shape[0])
self.row_unknowns = np.zeros(shape[1])
self.unknowns = 0
rind = data.domain.index(row_variable)
cind = data.domain.index(col_variable)
for row in data:
rval, cval = row[rind], row[cind]
w = row.weight
if np.isnan(rval) and np.isnan(cval):
self.unknowns += w
elif np.isnan(rval):
self.row_unknowns[int(cval)] += w
elif np.isnan(cval):
self.col_unknowns[int(rval)] += w
else:
self[int(rval), int(cval)] += w
self.row_variable = row_variable
self.col_variable = col_variable
return self
